A New RSSI-based Centroid Localization Algorithm by Use of Virtual Reference Tags

Transcription

1 INFOCOMP 0 : The Thrd Internatonal Conference on Adanced Communcatons and Computaton A New -based Centrod Localzaton Algorthm b Use of Vrtual Reference Tags Hua Long,, Joan Lu, Qang Xu School of Computng and Engneerng, Unerst of Huddersfeld, Huddersfeld, UK Facult of Informaton and Automaton, Kunmng Unerst of Scence and Technolog, Kunmng, Yunnan, P. R. Chna @.com; {j.lu,.u}@hud.ac.uk Qan Le Facult of Informaton and Automaton, Kunmng Unerst of Scence and Technolog, Kunmng, Yunnan, P. R. Chna @.com Abstract A good desgn of node locaton s crtcal for effcent and effecte wreless communcatons. Ths paper presents an mproed algorthm, n order to sole the low localzaton accurac caused b tradtonal centrod algorthm. The mproed algorthm combned wth VIRE sstem and tradtonal centrod algorthm. The VIRE algorthm s ntroduced and the sgnal propagaton model s utlzed to construct rtual reference tags n the locaton area. Smulaton shows that ths further deeloped algorthm has further mproed the accurac of postonng up to 5.% compared to the tradtonal centrod algorthm. It s concluded that ths algorthm can further mproe the locatng accurac n comparson wth the orgnal centrod algorthm. Kewords VIRE algorthm; Vrtual reference tags; RSS; Centrod algorthm. I. INTROUCTION Node localzaton s one of the core technologes n Wreless Sensor Network WSN []. Currentl, the localzaton algorthm of wreless sensor network can be dded nto two categores: the dstance-ndependence [] and dstance-dependence []. The common postonng algorthms of dstancendependence are centrod algorthm and Appromate Pont- In-trangulaton Test APIT [, 5]. Bulusu et al. [] ntroduced a node postonng technolog-centrod algorthm that has a lower computatonal complet than that of the trlateraton [5]. The postonng algorthm, such as Receed Sgnal Strength Indcator, of dstance-dependence has better accurac, but the sural of the network s stronger when the unknown node has not had a hgh postonng accurac. In order to mproe the accurac of centrod localzaton, ths paper proposes a modfed algorthm smlar to the Vrtual Reference Elmnaton VIRE algorthm [6]. Ths modfed algorthm uses the sgnal propagaton model [7] to construct the Receed Sgnal Strength Indcator [8] nformaton of rtual reference tags n locaton area and locates the object usng tradtonal centrod algorthm. The VIRE realzes accurate poston usng a lnear nterpolaton algorthm wth addng a grd rtual reference tags based on the LANMARC [9] algorthm. To reduce the redundant calculaton of VIRE algorthm, Sh et al. [0] used Subregon Selecton SRS mechansm to reduce the nterpolaton area, and a nonlnear nterpolaton algorthm to calculate the alues of rtual reference tags. Neertheless, the locaton accurac s affected b ddng areas [0]. To sole the aboe problem, ths paper proposes an mproed algorthm that combnes wth VIRE sstem and the orgnal centrod algorthm accordng to the VIRE algorthm and the nformaton of rtual reference tags that are establshed b sgnal transmsson model n the locaton area. The smulaton results show that the algorthm can further mproe the locatng accurac wthout an addtonal hardware oerhead n comparson wth the orgnal centrod algorthm. The rest of the paper wll be organzed as follows. Secton II wll descrbe locaton, VIRE dsposng and transmsson model of wreless sgnal. Secton III wll descrbe the detaled research approach. Secton IV wll descrbe the eperments and error analss. Fnall, Secton V wll draw the concluson and future work. II. RELATE WORK A. Rangng methods Accordng to the phscal medum, there are four popularl rangng methods n the sgnal communcatons of sensor sstems: the Receed Sgnal Strength Indcator [8], the TOA Tme Of Arral [], the TOA Tme fference Of Arral [], and the AOA Angle Of Arral []. The performance comparson of four rangng methods s shown n Table I. Coprght c IARIA, 0. ISBN:

2 INFOCOMP 0 : The Thrd Internatonal Conference on Adanced Communcatons and Computaton TABLE I. THE COMPARISON OF RANGING METHO BASE ON ISTANCE The TOA algorthm reures GPS-asssted eupment, whch s not used n the ndoor enronment [6]. Ultrasonc transmsson dstance s lmted, whch s usuall less than 0 meters, affected b angle, temperature, and obstructons [7]. The AOA s relable onl n the lne of sght sgnal domnant crcumstances [8]. It follows that these lmtatons of three postonng methods lead to an etreme dffcult n dealng wth the comple ndoor enronments. The -based rangng method has the adantages of long transmsson dstance, lower cost, lower power consumpton, eas node hardware, small sze, and lght weght [9]. Howeer, the postonng accurac of based methods s low, as t can onl be used as a rough rangng method [9]. Thus, ts postonng accurac must be mproed b an adjunct wa. B. Transmsson model of wreless sgnal The Receed Sgnal Strength Indctor conerts the loss durng the process of transmsson accordng to the sgnal ntenst of transmttng and node receed []. In the ndoor locaton, the rangng sgnal propagaton model can be smplfed as n consderaton of enronmental, cost, locaton accurac reurements and other factors: 0n log 0 d k wheren, n s the sgnal attenuaton factor that s from to ; d s the dstance of reference node and object node; k s the measured of one meter, wth the unt of dbm. III. RESEARCH APPROACH The approach s desgned n two phases: frst, the sstem structure and the algorthm based on rtual tags; second, the realzaton of algorthm n VIRE sstem. A. Sstem structure and algorthm based on rtual tags Constructon of VIRE sstem VIRE dsposng Vrtual Reference Elmnaton VIRE s ntroduced n [6]. In the VIRE, nstead of deplong man real reference RFI tags, the concept of rtual reference tags s proded as denser reference coerage n the sensng areas. The dstrbuton of VIRE reader Rand reference tags, T, a T, b T and c T are shown n Fgure wth relatonshp, d a, b,c,d N. The object tags are n the range of specfed area. The core of VIRE method s related to reference tags as a unt grd: NN, shown n Fgure. Fgure. The dstrbuton of VIRE readers and tags Fgure. The dstrbuton of rtual reference tags Under the condton of known alues of reference tags poston and denst, the feld strength of rtual reference tags n both horzontal and ertcal drectons s calculated b and : E R t, b E u, R E a N, b E a, b R R ER a, b t N E a, b N E a, b R R ER a, u N wheren, E s alue of reference tag located at R the coordnate, j for the R reader. The alues of parameters are a / N, b j / N, t s known and located at as as the nteral of object rtual reference tag, and u s known and located at as as the nteral of object rtual reference tag. Under the condtons of known alues of the feld ntenst of reference tags and rtual reference tags, LANMARC algorthm s normall used to select a certan threshold for solng the nearest reference tag weghts, n order to obtan postonng results [0]. Accordng to the VIRE sstem, ths paper ges a further deelopment of the postonng method. Coprght c IARIA, 0. ISBN:

3 INFOCOMP 0 : The Thrd Internatonal Conference on Adanced Communcatons and Computaton Fgure. Process of selecton of rtual reference tags Fgure. The defnton of rtual reference tags area The postonng sstem s for a two-dmensonal plane. The locaton area has a sze of S T, where reference tags are the four ertces of the regon, and the coordnates are T 0,0, T 0, T, T S, T, T 0, S. Four readers are placed on those ertces, as shown n Fgure. Assumng the Coordnate of rtual reference tag s, j, the Eucldean dstance between and reference tags T are: PT - - S -T PT T PT S PT - Accordng to aboe, the tables are establshed for the four reference tags shown n Table II below. TABLE II. VIRTUAL REFERENCE TAGS FOR REFERENCE TAGS stance of rtual reference alue of rtual reference Tags to the reference tags Tags relate to the reference tags PT PT PT PT There are N data n the table. PT PT PT PT The selecton of rtual tags Q, An object node s located at, the alues wll be read b readers at four corners. A comparson was carred out wth the rtual reference tags from readers. The rtual reference tags that are nearest dstance between Q, are selected accordng to the followng algorsm. IF >0 Q THEN for to N - do Value Q ELSE Return mn Value Return end End The four rtual reference tags are selected n Fgure as below. V,, V,, V,, V, B. The realzaton of algorthm n VIRE sstem V Tradtonal centrod algorthm Accordng to the tradtonal centrod algorthm, the four rtual reference tags are selected to form an enclosed area that s the estmated area of the object node []. B 5, the estmate coordnates of object node can be calculated. 5 Coprght c IARIA, 0. ISBN:

4 Fgure 5. The schematc dagram of the renewed centrod algorthm Renewed centrod algorthm Through narrowng down the testng area, a ke mproement of centrod algorthm n postonng accurac s descrbed as follows. Accordng to the geometrc relatonshp, n Fgure 5, the coordnates: e, f, g, h, and o, are defned n the followng relatonshps of 6, 7, 8, 9 and 0: e e 6 f f 7 e g 8 h h 9 o o 0 The process of the renewed algorthm s as shown n Table III. TABLE III. PROCESS OF THE RENEWE ALGORITHM 5 Coprght c IARIA, 0. ISBN: INFOCOMP 0 : The Thrd Internatonal Conference on Adanced Communcatons and Computaton

5 INFOCOMP 0 : The Thrd Internatonal Conference on Adanced Communcatons and Computaton IV. THE SIMULATION AN ERROR ANALYSIS In the smulaton eperment, Matlab7.6.0 software s used. There are two ndcators of sze set up and the postonng accurac s set as the ealuaton crtera, when the postonng precson, scale, network coerage rate, and power consumpton are often used as the ealuaton nde n the WSN. In the smulaton eperment, the WSN area s set that the S and T are eual to 90. The reference tags are located at the four ertces of ths area and unforml dstrbuted. The postonng error of the object nodes s defned n : - - real real AerageError s Wheren, s s the tmes of the Smulaton. A. Searchng the optmum number of the rtual reference tags As shown n Table IV, ten object nodes are from the reader R, where s shown as,,,. TABLE IV OBJECT NOES S When the number of the rtual reference tags ncreases, the error n comparson wth the two algorthms are llustrated n Fgures 6 and 7. Fgure 6 shows that there s no obous change between the two algorthms, f the number of the rtual reference tags s aboe and the 90/N s greater than 0. The best performance shows that the number of the rtual reference tags s 9 when 90/N s eual to 6. It follows that the renewed centrod algorthm could further mproe the accurac compared wth the tradtonal centrod algorthm. Fgure 6. The dfference between two algorthms Coprght c IARIA, 0. ISBN:

6 INFOCOMP 0 : The Thrd Internatonal Conference on Adanced Communcatons and Computaton Fgure 7. The error comparson of two algorthms when the number of the rtual reference tags ncreases B. Comparson of the two algorthms when the number of rtual reference tags s 9 Fgure 8 shows that the results are obtaned b 5 from the test for 00 tmes. It s clear that the renewed centrod algorthm gets a further mproed localzaton accurac compared wth the tradtonal one. C. Error analss For the rtual reference tags, the ma not be able to choose four ones near the object node, or four rtual reference tags are repeatedl chosen, as t ma lead to ecesse errors. In the rado propagaton model, the sgnal s dsturbed b the eternal enronment nstead of beng transmtted n free space. Through repeated measurements and the statstcal aerage alue, some error can be reduced, although the further mproement s stll needed. Fgure 8. The comparson of the two algorthms V. CONCLUSION AN FUTURE WORK A. Acheements One of the hot topcs n WSN s localzaton of the object node. The paper proposes an mproed centrod algorthm combned wth the VIRE sstem. The algorthm s smple. The postonng accurac of the object nodes has been mproed sgnfcantl n comparson wth the tradtonal centrod algorthm. Ths new algorthm can be used to sole the problem of the low localzaton accurac of the tradtonal centrod algorthm. Coprght c IARIA, 0. ISBN:

7 INFOCOMP 0 : The Thrd Internatonal Conference on Adanced Communcatons and Computaton B. Future work Howeer, there are stll some shortcomngs estng n the postonng methods. On one hand, the sgnal propagaton model applcable to non-free-space condtons needs further research. On the other hand, the selecton methods of the rtual reference tags need further optmzaton so as to further mproe the postonng accurac. Acknowledgement: authors thank r Sawsaa s feedback and proof readng for the fnal erson of the paper. REFERENCES [] S. K. Low, H. A. Nguen, and H. Guo, Optmzaton of sensor node locatons n a WSN[C]//Fourth Internatonal Conference on Natural Computaton. 008, pp [] G. Mao, B. Fdan, and B.. O. Anderson, Wreless sensor network localzaton technues, Comput. Netw., ol. 5, pp , 007. [] C. Lu, H. Jang, and. Wang, Range-based node localzaton algorthm for wreless sensor network usng Untar Matr pencl[c]//test and Measurement, 009, ICTM'09, Internatonal Conference on. IEEE, 009, ol., pp. 8-. [] N. Bulusu, J. Hedemann, and. Estrn, GPS-less lowcost outdoor localzaton for er small deces [J], Personal Communcatons, IEEE, 000, ol. 75, pp. 8-. [5]. Moore, J. Leonard, and. Rus, Robust strbuted Network Localzaton wth Nos Range Measurements[C]. Proceedngs of the nd Internatonal Conference on Embedded Networked Sensor Sstems, Baltmore, USA, 00: [6] Z. Yang and L. Yunhao, VIRE: Acte RFI-based localzaton usng rtual reference elmnaton[c], XAn, 007 Internatonal Conference on Parallel Processng, 007, pp [7] M. Srastaa,. Culler, and. Estrn, Guest edtors' ntroducton: oerew of sensor networks [J], Computer, 00, ol. 78, pp. -9. [8] L. Grod, V. Bchosk, J. Elson, and. Estrn, Locatng Tn Sensors n Tme and Space: A Case Stud, [C]//Proceedngs of the 00 IEEE Internatonal Conference on Computer esgn: VLSI n Computers and Processors. Freburg, IEEE, 00, pp. -9. [9] M. N. Lonel, Y. H. Lu, and Y. C. Lau, LANMARC: Indoor locaton sensng usng acte RFI [J], Wreless Networks, 00, ol. 06, pp. 70l-7l0. [0] W. Sh, L. Lu, Y. Ju, G. Yan, and G. An Effcent Indoor Locaton Algorthm Based on RFI Technolog, [C]//Wreless Communcatons Networkng and Moble Computng WCOM, 00 6th Internatonal Conference on. IEEE, 00, pp. -5. [] J. Caffer, A new approach to the geometr of TOA locaton [C]//IEEE VTC000 Fall September 8, Boston, USA, 000, pp [] J. Yoon, J W. Km, W. Y. Lee, and. S. Eom, A TOA-based localzaton usng precse tmesnchronzaton[c], Adanced Communcaton Technolog ICACT, 0 th Internatonal Conference on, IEEE, 0, pp [] J. Zhou, H. Zhang, and L. Mo, Two-dmenson localzaton of passe RFI tags usng AOA estmaton, [C]//Instrumentaton and Measurement Technolog Conference IMTC, 0 IEEE, 0, pp. -5. [] L.M. Sun, J.Z. L, and Y. Chen, WSNs [M], Tsnghua Unerst Press, 005, pp [5] H. Sh, A new weghted centrod localzaton algorthm based on, [C]//Informaton and Automaton ICIA, 0 Internatonal Conference on. IEEE, 0, pp. 7-. [6] N. Saedeh and O. M. Raza, GPS-less ndoor constructon locaton sensng, orgnal research artcle, Automaton n Constructon, ol. 8, ecember 0, pp. 8-6 [7] X. Guangzhen, P. Yang, and L. He, Estmaton of dffracton effect n ultrasonc attenuaton b throughtransmsson substtuton technue, orgnal research artcle, Ultrasoncs, Aprl 0, ol. 5, Issue, pp [8] M. akkak, A. Nakb, B. aach, P. Sarr, and J. Lemone, Indoor localzaton method based on RTT and AOA usng coordnates clusterng, orgnal research artcle, Computer Networks, June 0, ol. 55, Issue 8, pp [9] W. Huanhuan, J. Wang, and R. Lu, A Noel Rangng Method Based on, orgnal research artcle, Energ Proceda, 0, ol., pp Coprght c IARIA, 0. ISBN: